A Nesterov-Like Gradient Tracking Algorithm for Distributed Optimization Over Directed Networks

被引:47
作者
Lu, Qingguo [1 ]
Liao, Xiaofeng [2 ]
Li, Huaqing [1 ]
Huang, Tingwen [3 ]
机构
[1] Southwest Univ, Coll Elect & Informat Engn, Chongqing Key Lab Nonlinear Circuits & Intelligen, Chongqing 400715, Peoples R China
[2] Chongqing Univ, Coll Comp, Chongqing 400044, Peoples R China
[3] Texas A&M Univ Qatar, Sci Program, Doha, Qatar
来源
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS | 2021年 / 51卷 / 10期
基金
中国国家自然科学基金;
关键词
Convergence; Cost function; Convex functions; Acceleration; Delays; Information processing; Directed network; distributed convex optimization; gradient tracking; linear convergence; Nesterov-like algorithm; LINEAR MULTIAGENT SYSTEMS; CONVERGENCE; CONSENSUS; GRAPHS; ADMM;
D O I
10.1109/TSMC.2019.2960770
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In this article, we concentrate on dealing with the distributed optimization problem over a directed network, where each unit possesses its own convex cost function and the principal target is to minimize a global cost function (formulated by the average of all local cost functions) while obeying the network connectivity structure. Most of the existing methods, such as push-sum strategy, have eliminated the unbalancedness induced by the directed network via utilizing column-stochastic weights, which may be infeasible if the distributed implementation requires each unit to gain access to (at least) its out-degree information. In contrast, to be suitable for the directed networks with row-stochastic weights, we propose a new directed distributed Nesterov-like gradient tracking algorithm, named as D-DNGT, that incorporates the gradient tracking into the distributed Nesterov method with momentum terms and employs nonuniform step-sizes. D-DNGT extends a number of outstanding consensus algorithms over strongly connected directed networks. The implementation of D-DNGT is straightforward if each unit locally chooses a suitable step-size and privately regulates the weights on information that acquires from in-neighbors. If the largest step-size and the maximum momentum coefficient are positive and small sufficiently, we can prove that D-DNGT converges linearly to the optimal solution provided that the cost functions are smooth and strongly convex. We provide numerical experiments to confirm the findings in this article and contrast D-DNGT with recently proposed distributed optimization approaches.
引用
收藏
页码:6258 / 6270
页数:13
相关论文
共 61 条
[11]   Differentially Private Distributed Online Learning [J].
Li, Chencheng ;
Zhou, Pan ;
Xiong, Li ;
Wang, Qian ;
Wang, Ting .
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2018, 30 (08) :1440-1453
[12]   Accelerated Convergence Algorithm for Distributed Constrained Optimization under Time-Varying General Directed Graphs [J].
Li, Huaqing ;
Lu, Qingguo ;
Liao, Xiaofeng ;
Huang, Tingwen .
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS, 2020, 50 (07) :2612-2622
[13]   Convergence Analysis of a Distributed Optimization Algorithm with a General Unbalanced Directed Communication Network [J].
Li, Huaqing ;
Lu, Qingguo ;
Huang, Tingwen .
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING, 2019, 6 (03) :237-248
[14]   Event-Triggered Communication and Data Rate Constraint for Distributed Optimization of Multiagent Systems [J].
Li, Huaqing ;
Liu, Shuai ;
Soh, Yeng Chai ;
Xie, Lihua .
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS, 2018, 48 (11) :1908-1919
[15]   Cooperative Optimization of Dual Multiagent System for Optimal Resource Allocation [J].
Li, Kaixuan ;
Liu, Qingshan ;
Yang, Shaofu ;
Cao, Jinde ;
Lu, Guoping .
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS, 2020, 50 (11) :4676-4687
[16]  
Li Z., 2016, ARXIV PREPRINT ARXIV
[17]   Performing linear convergence for distributed constrained optimisation over time-varying directed unbalanced networks [J].
Lu, Qingguo ;
Li, Huaqing ;
Wang, Zheng ;
Han, Qi ;
Ge, Wei .
IET CONTROL THEORY AND APPLICATIONS, 2019, 13 (17) :2800-2810
[18]   Geometrical convergence rate for distributed optimization with time-varying directed graphs and uncoordinated step-sizes [J].
Lu, Qingguo ;
Li, Huaqing ;
Xia, Dawen .
INFORMATION SCIENCES, 2018, 422 :516-530
[19]  
Maros M., 2018, ARXIV PREPRINT ARXIV
[20]  
Maros M, 2018, IEEE DECIS CONTR P, P6520, DOI 10.1109/CDC.2018.8619626